olivierlambert/calrs v2.0.0-rc.2

2.0.0-rc.2

on GitHub

Second release candidate of 2.0, with one change: the bundled timezone data. calrs carried IANA tzdata 2026b, and four releases since then move zones to permanent offsets, most of them already in effect, so times in Alberta, the Northwest Territories, Manitoba, Morocco and Cuba were an hour off. Upgrade from rc.1 if you, your guests or your calendar events use one of those zones. Synced calendar events are read correctly again as soon as you upgrade; bookings and time off saved while the data was stale keep the instant they were saved with, so check those made in these zones since the change. CI now fails while the bundled data is older than IANA's latest, so this cannot go stale silently again. The same data ships for the 1.23 line as 1.23.4. No migrations since rc.1.

Docker: this version is published as :2.0.0-rc.2 and :next. :latest stays on the 1.23 line (now 1.23.4) until 2.0 is final.

Fixed

• Timezone data updated to IANA 2026f (#285) - calrs bundled tzdata 2026b, and four releases since then move zones to permanent offsets, most already in effect: Morocco +00 from 2026-09-20, Manitoba (America/Winnipeg) -05 from 2026-10-31, Alberta (America/Edmonton) and the Northwest Territories -06 from 2026-11-01, and Cuba staying on -04 until 2027-11-07. For hosts, guests or calendar events in those zones, every time (slots, busy times, emails, invites) was an hour off. Tests pin each change. Bookings and time off saved while the data was stale keep the instant they were saved with.

Internal

• CI fails when the bundled tz data is stale (#285) - scripts/check-tzdata.sh compares vendor/chrono-tz/tz with IANA's latest release. CI runs it on every push and pull request, and a daily workflow opens an issue when IANA publishes, so a new release is noticed before a zone's clocks change.
• A cross-check of the bundled tz data against IANA's tools (#285) - vendor/chrono-tz/crosscheck.py and the opt-in test tzdata_matches_zdump compare what calrs reads with zic/zdump for every zone, at each transition from 1800 and monthly to 2040. It caught two regressions in the new data before release, where the parser calrs builds with ignores a time suffix (Alaska's 1867 changeover, Dublin's 1925 return to GMT); the vendored copy writes both in wall time.